The effects of topological features on convolutional neural networks—an explanatory analysis via Grad-CAM
Topological data analysis (TDA) characterizes the global structure of data based on topological invariants such as persistent homology, whereas convolutional neural networks (CNNs) are capable of characterizing local features in the global structure of the data. In contrast, a combined model of TDA...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IOP Publishing
2023-01-01
|
| Series: | Machine Learning: Science and Technology |
| Subjects: | |
| Online Access: | https://doi.org/10.1088/2632-2153/ace6f3 |
| _version_ | 1827866094494285824 |
|---|---|
| author | Dongjin Lee Seong-Heon Lee Jae-Hun Jung |
| author_facet | Dongjin Lee Seong-Heon Lee Jae-Hun Jung |
| author_sort | Dongjin Lee |
| collection | DOAJ |
| description | Topological data analysis (TDA) characterizes the global structure of data based on topological invariants such as persistent homology, whereas convolutional neural networks (CNNs) are capable of characterizing local features in the global structure of the data. In contrast, a combined model of TDA and CNN, a family of multimodal networks, simultaneously takes the image and the corresponding topological features as the input to the network for classification, thereby significantly improving the performance of a single CNN. This innovative approach has been recently successful in various applications. However, there is a lack of explanation regarding how and why topological signatures, when combined with a CNN, improve discriminative power. In this paper, we use persistent homology to compute topological features and subsequently demonstrate both qualitatively and quantitatively the effects of topological signatures on a CNN model, for which the Grad-CAM analysis of multimodal networks and topological inverse image map are proposed and appropriately utilized. For experimental validation, we utilize two famous datasets: the transient versus bogus image dataset and the HAM10000 dataset. Using Grad-CAM analysis of multimodal networks, we demonstrate that topological features enforce the image network of a CNN to focus more on significant and meaningful regions across images rather than task-irrelevant artifacts such as background noise and texture. |
| first_indexed | 2024-03-12T14:59:04Z |
| format | Article |
| id | doaj.art-feea9e587d5e469394ce9782b9f5be7c |
| institution | Directory Open Access Journal |
| issn | 2632-2153 |
| language | English |
| last_indexed | 2024-03-12T14:59:04Z |
| publishDate | 2023-01-01 |
| publisher | IOP Publishing |
| record_format | Article |
| series | Machine Learning: Science and Technology |
| spelling | doaj.art-feea9e587d5e469394ce9782b9f5be7c2023-08-14T10:55:53ZengIOP PublishingMachine Learning: Science and Technology2632-21532023-01-014303501910.1088/2632-2153/ace6f3The effects of topological features on convolutional neural networks—an explanatory analysis via Grad-CAMDongjin Lee0https://orcid.org/0009-0003-0777-6717Seong-Heon Lee1https://orcid.org/0009-0003-3902-5678Jae-Hun Jung2https://orcid.org/0000-0001-7923-0674Graduate School of Artificial Intelligence, Pohang University of Science and Technology (POSTECH) , Pohang 37673, Republic of KoreaGraduate School of Artificial Intelligence, Pohang University of Science and Technology (POSTECH) , Pohang 37673, Republic of KoreaGraduate School of Artificial Intelligence, Pohang University of Science and Technology (POSTECH) , Pohang 37673, Republic of Korea; Department of Mathematics, Pohang University of Science and Technology (POSTECH) , Pohang 37673, Republic of KoreaTopological data analysis (TDA) characterizes the global structure of data based on topological invariants such as persistent homology, whereas convolutional neural networks (CNNs) are capable of characterizing local features in the global structure of the data. In contrast, a combined model of TDA and CNN, a family of multimodal networks, simultaneously takes the image and the corresponding topological features as the input to the network for classification, thereby significantly improving the performance of a single CNN. This innovative approach has been recently successful in various applications. However, there is a lack of explanation regarding how and why topological signatures, when combined with a CNN, improve discriminative power. In this paper, we use persistent homology to compute topological features and subsequently demonstrate both qualitatively and quantitatively the effects of topological signatures on a CNN model, for which the Grad-CAM analysis of multimodal networks and topological inverse image map are proposed and appropriately utilized. For experimental validation, we utilize two famous datasets: the transient versus bogus image dataset and the HAM10000 dataset. Using Grad-CAM analysis of multimodal networks, we demonstrate that topological features enforce the image network of a CNN to focus more on significant and meaningful regions across images rather than task-irrelevant artifacts such as background noise and texture.https://doi.org/10.1088/2632-2153/ace6f3convolutional neural networktopological data analysisGrad-CAM analysisCNN-TDA Nettransient vs bogus classificationpigmented skin classification |
| spellingShingle | Dongjin Lee Seong-Heon Lee Jae-Hun Jung The effects of topological features on convolutional neural networks—an explanatory analysis via Grad-CAM Machine Learning: Science and Technology convolutional neural network topological data analysis Grad-CAM analysis CNN-TDA Net transient vs bogus classification pigmented skin classification |
| title | The effects of topological features on convolutional neural networks—an explanatory analysis via Grad-CAM |
| title_full | The effects of topological features on convolutional neural networks—an explanatory analysis via Grad-CAM |
| title_fullStr | The effects of topological features on convolutional neural networks—an explanatory analysis via Grad-CAM |
| title_full_unstemmed | The effects of topological features on convolutional neural networks—an explanatory analysis via Grad-CAM |
| title_short | The effects of topological features on convolutional neural networks—an explanatory analysis via Grad-CAM |
| title_sort | effects of topological features on convolutional neural networks an explanatory analysis via grad cam |
| topic | convolutional neural network topological data analysis Grad-CAM analysis CNN-TDA Net transient vs bogus classification pigmented skin classification |
| url | https://doi.org/10.1088/2632-2153/ace6f3 |
| work_keys_str_mv | AT dongjinlee theeffectsoftopologicalfeaturesonconvolutionalneuralnetworksanexplanatoryanalysisviagradcam AT seongheonlee theeffectsoftopologicalfeaturesonconvolutionalneuralnetworksanexplanatoryanalysisviagradcam AT jaehunjung theeffectsoftopologicalfeaturesonconvolutionalneuralnetworksanexplanatoryanalysisviagradcam AT dongjinlee effectsoftopologicalfeaturesonconvolutionalneuralnetworksanexplanatoryanalysisviagradcam AT seongheonlee effectsoftopologicalfeaturesonconvolutionalneuralnetworksanexplanatoryanalysisviagradcam AT jaehunjung effectsoftopologicalfeaturesonconvolutionalneuralnetworksanexplanatoryanalysisviagradcam |